The original two-parameter Weibull distribution used for rolling element bearing fatigue tends to greatly underestimate life at high levels of reliability. This fact has been proven for through hardened ball, cylindrical and spherical roller bearings, as well as linear ball bearings, by other researchers. However, to date this has not been done with tapered roller bearings (TRB) or case carburized materials, and as such this study was conducted. First, the three-parameter Weibull distribution was utilized to create a mathematical model, and statistical data analysis methods were put into place. This algorithm was then investigated as to its ability to discern the shape of the reliability distribution using known, numerically generated, data sets for two and three-parameter Weibull distributions. After validation, an experimental data set of 9702 TRB’s, 98% of which were case carburized, was collected. Using the developed algorithm on this data set, the overall RMS error was reduced from 26.0% for the standard, two-parameter to 12.2% for the three-parameter Wiebull distribution. Also, the error at 99.9% reliability was reduced from 95.8% to 37%. However, as the results within varied from previously published values at high reliabilities, there is likely a difference in the underlying population and/or dependency on the statistical and mathematical methods utilized. Therefore, more investigation should be conducted in this area to identify the underlying variables and their effects on the results.
